Dear Statalisters,
I have been trying to run a model explaining if managers react to problems
they encounter with their suppliers. Since I expect a selection problem to
occur (i.e. managers try to achieve the best outcomes by choosing one of the
two possible strategies) a two-staged Heckman approach seems appropriate.
Therefore I have constructed a selection equation based on some aspects of
the severity of problems they encounter in the transaction together with
aspects of the buyer-seller contract. Subsequently I have controlled for
each of these exogenous variables plus one instrument in the outcome
equation as follows:
heckman outcome additionalvar selectionvars, twostep select(strategy =
selectionvars) rhosigma
The outcomes using this approach indeed suggest a specification issue as the
outcomes between a Heckman and OLS estimation differ substantially (see
below with the original variable names corresponding to the above). However,
this would lead me to expect that the inverse Mills ratio is also
significant but this is not the case. How could this be? All the examples
that I have seen on the Internet on Heckman show that substantial parameter
differences between Heckman and OLS relate to significant Mills ratios. Is
there anything I have done wrong?
With kind regards,
Erik
-----------------
heckman v5_12b v5_2 exposttc rights service prodpric recourse, twostep
select(v5_5bf = exposttc rights service prodpric recourse) rhosigma
Heckman selection model -- two-step estimates Number of obs =
928
(regression model with sample selection) Censored obs =
737
Uncensored obs =
191
Wald chi2(11) =
227.39
Prob > chi2 =
0.0000
----------------------------------------------------------------------------
--
| Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+----------------------------------------------------------
v5_12b |
v5_2 | .0971938 .2696069 0.36 0.718 -.431226
.6256136
exposttc | -.1511838 .0830238 -1.82 0.069 -.3139074
.0115397
rights | .0937918 .1002321 0.94 0.349 -.1026595
.2902432
service | -.0115216 .1530307 -0.08 0.940 -.3114562
.288413
prodpric | -.109601 .0761067 -1.44 0.150 -.2587675
.0395654
recourse | -.0503901 .125504 -0.40 0.688 -.2963735
.1955933
_cons | 10.51196 3.910816 2.69 0.007 2.846904
18.17702
-------------+----------------------------------------------------------
v5_5bf |
exposttc | .0993788 .0069866 14.22 0.000 .0856854
.1130722
rights | -.0959504 .0315393 -3.04 0.002 -.1577663
-.0341346
service | .1273116 .0566456 2.25 0.025 .0162883
.238335
prodpric | .023815 .0387175 0.62 0.538 -.05207
.0997
recourse | .1158447 .0425291 2.72 0.006 .0324892
.1992002
_cons | -3.228235 .2235185 -14.44 0.000 -3.666323
-2.790147
-------------+----------------------------------------------------------
mills |
lambda | -.2353879 1.312908 -0.18 0.858 -2.80864
2.337864
-------------+----------------------------------------------------------
rho | -0.17707
sigma | 1.3293293
lambda | -.23538794 1.312908
----------------------------------------------------------------------------
--
reg v5_12b v5_2 exposttc rights service prodpric recourse
Source | SS df MS Number of obs =
915
-------------+------------------------------ F( 6, 908) =
95.83
Model | 726.998251 6 121.166375 Prob > F =
0.0000
Residual | 1148.04498 908 1.26436672 R-squared =
0.3877
-------------+------------------------------ Adj R-squared =
0.3837
Total | 1875.04324 914 2.05146962 Root MSE =
1.1244
----------------------------------------------------------------------------
--
v5_12b | Coef. Std. Err. t P>|t| [95% Conf.
Interval]
-------------+----------------------------------------------------------
v5_2 | -.1803181 .0917545 -1.97 0.050 -.3603936
-.0002425
exposttc | -.1082438 .0056015 -19.32 0.000 -.1192371
-.0972505
rights | .0120973 .0219006 0.55 0.581 -.0308845
.055079
service | .0079959 .0387689 0.21 0.837 -.0680912
.084083
prodpric | -.0042016 .0250155 -0.17 0.867 -.0532966
.0448933
recourse | .01187 .0307798 0.39 0.700 -.0485377
.0722778
_cons | 9.398066 .1371511 68.52 0.000 9.128896
9.667236
----------------------------------------------------------------------------
--
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/